A new topology for building a low-cost DIY HV power supply ... 30kV/80W for $25.

[Rich Feldman]

Nice work there Chris, on a great idea. I'm looking forward to seeing more.
-Rich

[Chris Bradley]

Here's pictures that will help you make sense of the 'bolt' for shorting the shunt capacitor out of the circuit, that I've described and that appears in the photos of my prototypes above.

First 4 image samples show;
- an inverter from the top,
- one that has had the aluminium stripped off a sample that has had a poor and incomplete potting (this is easily done by hand, just cut off the connectors at the wire ends and bend the can off), and in this example I have soldered a wire onto the pad that needs to be shorted to ground,
- an inverter with a hole drilled 18mm to the left of the end with the output (white) wires, and 5.5mm up from the bottom, per orientation shown,
- an x-ray of an inverter, the shunt capacitor being shown in the lower right of the picture (which is what we need to short out).

So the steps I followed to modify my inverters are;
1) Drill a 2.5mm hole 18mm from the right and 5.5mm up, with the output side to the right. If the label is stuck generally central, then this is just between, and slightly high above, the 'g' and 'e' of 'verlangern'.
i) drill so as to just puncture through the Al case,
ii) stop and inspect the hole - if black material has come out with the drilling, there's potting there, go to step 1(iii). If not, inspect with a torch and probe, e.g. with a fine screwdriver - is there any resistance to putting in the screwdriver in (note - you may come across a piece of insulating paper in there), or can you reach down with it to, or see, a soldered pad in the hole? If so, you may prefer to strip the Al case off, as shown in the photos, and solder to the pad directly. If you don't want to do it that way, go to step 2.
iii) keep drilling further down, the drill will cut easily through the potting until you feel some resistance - which should be the board and the solder pad. Drill just an extra mm beyond that, and a little bit of metal [solder] should come out of the hole by the drill flutes. Stop there!
2) Tap the inverter on the workbench, upside down, so debris drops out of the hole. Take a 3mm machine bolt. It will self-tap into the 2.5mm hole if you hold it carefully and perpendicular as it bites the Al. Screw it in until you feel resistance as it bottoms out in the hole. Don't over tighten, it'll easily strip the self-tapped thread in that Al case.
3) Measure the resistance between the case and each of the white output lines. One should now read ~146 Ohms (the upper lead, as you look at these photos). The other will be DC isolated but you might be able to read the 50pF cap on it - cut that lead off.
4) If you don't read ~146Ohms on either lead then tighten the bolt up a little further. Keep tightening the bolt up a little until you read ~146Ohms. It should take around a half dozen turns of the screw to reach the solder point where you've shorted that connection to ground.
5) If (4) doesn't work out for you, maybe you needed to have drilled a little further. It should have been around 6~8mm deep. You could try drilling a bit deeper, though you've probably missed the spot due to some internal manufacturing anomaly (or it isn't like the ones I've done!).

If these steps have worked out, you should now have an inverter that looks like the next set of 4 photos, with one white lead with 146Ohms to ground (because the other side of the transformer is shorted to the case by the bolt) and the other lead is cut off.

If you have stripped the case off because there was no potting where you drilled, then solder to the point shown in the previous photos, and that should then be subsequently connected to the same connection as the black lead (input) when you rig up the circuit.

There is always the option that you just strip the case off the inverter in the first step and then drill or cut the potting away from the solder pad (if it is potted properly) and solder on to it. Your choice. But the Al package has seemed useful for screwing the inverter in place, and seems to help a little with LF shielding.

Please let me know if that is a good description for you, or if there are any more questions on preparing these inverters.

Also to say, once you have your inverters and know the V/I requirements you need for your supply, then let me know and I'll work with you to figure the best arrangement to achieve the output requirements. Max input requirements are 2-20V at N Amps (N = no of inverters) if you run all the input leads in parallel.

In the prototype you see above, I have arranged the six inverters in 3 pairs so that they can be connected all in parallel (20V x 6A) or in series (60V x 2A), which suits my bench supplies. You might have spotted a plastic divider between each pair - that's so the cases don't short to each other when they are wired in series.

...

[Ed Meserve]

Chris, really nice work, and thank-you for sharing. After seeing what you've done here, we are going to use your method for our power supply.

Did you take the x-rays of the inverter yourself?

regards,
Ed

[nicolas leboucher]

hello chris,

I ve received this week-end the inverters from

http://shop.wiltec.info/product_info.ph ... -40mm.html

and the model seems to be quiet different as yours.

as you can see on the first ans second pictures, the white wires now come out from the upper part of the alu pot side.

as you can see on the third picture, i stripped off the aluminium on a sample that has had a poor and incomplete potting.

white wire F is connected to pad A
white wire E is connected to pad B
a capacitor 50 pf have been measured between C and D

so for this model ( I ve got 10 of it) I will give within a few days new length measurements to drill the hole over pad D which is about 11mm from the right side ( without alu thickness) and 4mm from the bottom side (without alu thickness).

[Chris Bradley]

Hi Nicolas,

You're OK with these. I think the current batch has probably undergone a process modification to save a second (..?!) on manufacturing time by not giving any stress-relief routing to the output wires - they come straight out now, whereas on the ones I described above there is a loop that I presume was there originally to help avoid the wires getting pulled of the pads if the wire is strained. (Maybe - just a guess why that was like that in the first batch.)

So if you look at the comparison below, I think this is like yours and you have much the same, just that the wires are routed differently. That's all.

In regards the pad arrangement, if you take a look at the photo above where I have soldered a wire to a pad, that group of 4 solder points are the pads numbered 1,2,4&5 in the x-ray below, which are all electrically connected and are on the transformer side of the shunt cap. I chose to aim for pad 4, but any will do the task.

But looking at your inverter does look like there may be further differences to the board's pads. I can't quite tell. Looks like they have cut down the number of solder pads.

As long as you are measuring ~146Ohm (the ohmic resistance of the secondary) between your wire F and pad D then pad D would be as good as any to use. All you're after with this topology, or with any inverter, is to take one output of the transformer HV secondary to ground and the other you add your own shunt capacitor to (with a value high enough to isolate the output). Whatever way you do that, that's all it's about.

Well, in any case, I trust I've given the detail you need to determine which two pads are the capacitor's solders, and you can see now to short the transformer side of that cap to ground. Whether you do it with the can pulled off, or drill through as I've done, I think you'll be good to figure out for your own build now.

Good luck!

So, what output are you actually after, Nicholas? For your application of an SEM you should never experience a discharge, so you won't need those internal resistors I've shown in my prototype. Instead, all you'd need is a big-value ballast resistor just as a 'safety resistor', just in case of a fault mode. But if you want to ask further let me know what you want your outputs to be/do, and I'd be happy to advise what I think is the easiest arrangement to accomplish it. [If you don't need a big current (which should be the case for an SEM) then you can use one inverter per 2 or 3 stages instead... ] Or just feel free just to experiment for yourself!

...

[Chris Bradley]

Hi Ed, these were taken on a professional real-time x-ray microscope. I've pondered whether to attempt an x-ray with these power supplies, but it was just a 'pondering' at this stage. X-ray generation needs to be from a very small target for image focus, which should not be too difficult, but I've no idea on how to image the x-rays afterwards. I guess I'd need a 6" wide CCD!?

[Chris Bradley]

I had some issues with one of my supplies this weekend. I've ran the two prototype units you see above for a year with few problems. One I have had to replace a broken inverter on one occasion, and the other one twice when I was deliberately testing out sparking discharge robustness.

So, generally reliable in practice but occasionally the weakness I've described above shows up if it is subjected to a hard arcing load.

This weekend I was running some fairly routine tests but I had installed a 68k ballast resistor internally to the unit. I guess the current through it reached a level where some voltage potential had been reached and I heard a couple of arcs internally and found it was under-performing after that. The set-up was as per a negative polarity output, with the +ve end grounded, so the inverter at the -ve end took the most hammering. I replaced that but then the new one broke straight after too, which was odd (maybe a duff inverter - we're not talking 'expensive quality' here!)

Anyhow, I decided to implement one of the solutions presented in the patent application documents, above. You can see how a 'charge pump' arrangement can be set up in Figure 12. The photo below shows the practical implementation of this, on the lower stage (only).

I'm uncomfortable with the RC times in this, as these HV diodes have a trr of around 150ns. I'm going to try a range of higher shunt resistor values and I think they should be more comparable with the impedance (@40kHz) of the capacitors. I think it is better that they are in the order of a few 10's k. It is a drop of a couple of 100V at 5mA, but the stack itself doesn't suffer from insufficient volts - it is the current and arcing protection it needs. I think it is probably better to sacrifice more volts for safer tolerance to adverse loads. I'll do a few more experiments to see if there is some amount of ohmic shunt resistance that, essentially, fully (or mostly) protects it from arc discharging.

... If anyone has got any other good ideas on circuit implementations that can protect the inverters from external arc discharges, then I'd be interested to hear them ....


edit ... I've just now upped the shunt resistors to 68k and I can now strike repetitive 15kV sparks and no damage seems to have yet occurred. I'll test under continuous load to check that this amount of shunt resistance is not too great, but as the inverters themselves cannot really put much more than 5mA out, so I don't expect a continuous load to overheat the (3W) resistors.
...

[ab0032]

You have two sources for the current in the sparks,
1) discharge of the charge stored in the caps
2) current pumped in at D1 in your diagram.

Your 68k is quite big, I see two alternatives,

1) instead of putting one large resistor in at the end, you could try putting smaller resistors, perhaps 1ks into the diode chain between every two diodes and a final one that is smaller than 68k at the end of it.
That way the caps at the end will discharge quicker, but their charge is limited and after that the charge in the caps further down the cascade will discharge, but slower.

2) Ignore the charge in the caps, the energy stored in them is pretty small anyways (less than 1 Joule?), and simply limit the current going into D1 way at the bottom. This could pose a problem though, if you do not let new charge in but keep pumping, the lower end at D2 might reach a high potential too, in your case a high positive potential, as you keep pumping the electrons out the other end. Since the circuit is not made for this, it could damage something at the lower end of the voltage multiplier cascade. But because of this positive charge pumping could also break down or at least slow down.

I would be really interested to see how alternative 1) performs but also curious about the second.

[Chris Bradley]

Hi Alex. The issue isn't the current in the spark, per se - the output caps have no dv/dt issue. The issue is that a big dv/dt on the coupling caps will cause a sudden voltage rise at the inverters. I was thinking whether there is some suitable inductance to increase the impedance in a short rise time event, but as there may be a variety of external ballasting then I'm not sure if I can pick an inductance value that doesn't interfere or resonate with the coupling caps but does achieve the necessary impedance in a high dv/dt event.

So an ohmic impedance seems to be the easiest and best 'catch-all'.

68k is a drop of 340V at 5mA (a loss of 1.7W per resistor at max current). This means I'd be losing something under 10% of output power (at max power) to that much ohmic resistance, but it seems the safest and simplest way. It's really not much because there is an excess of volts from the inverters (they can pump 36kV, while the coupling caps are 30kV rated - I have 6kV I can 'lose' without impact of volts!) Bear in mind the 50pF coupling cap already presents an impedance of 80k at 40kHz, the issue is that their impedance for a high dv/dt drops substantially and therefore pulls the inverter around excessively (if you want to look at the maths of the electronics that way).

Also need to look at RC times - With 68k, the RC time is now around 3us, so as the voltage ramps up towards, say, 20kV on the end of the inverter then that protection diode (150ns trr) should be opening before it gets much above 2kV.

[ab0032]

Ahh you are one step ahead of me, I didnt think of the inverters. True.

The 68k could burn out if it is not 30kV proof which means you could get an arc there and hence also the current after the 68k is destroyed, so I think it would be better to use multiple resistors distributed around the circuit that can actually take the voltages they receive, eg between the diodes for example.

The thought of putting some mHenries into the output line also crossed my head, but I really dont know if that is such a good idea in the end... It would stop current from rising too quickly, but it would also keep it going when you might not want it any more leading to potential peaks in excess of the 30kV you might want.

[Chris Bradley]

The resistors might, indeed, see 30kV across them. But they are in series with a relatively paltry ~50pF, so they'll never have 30kV across them for very long.

This is the issue; for example, imagine there is an arc from the -HV (let's say it is at 30kV) to ground. The coupling capacitor at the lowest end will have one end oscillating (AC) around ground, and the other end will be being held at -30kV but as it is shorted so it leaps to 0V so the other end jumps up to 30kV. If the 68k resistor is in the way, then as that terminal jumps to 0V so it is the resistor that will have this 30kV across it, in that instant.

However it only has it momentarily because the current through it starts charging up the 50pF link capacitor. For a us, there will, indeed, be a ~half an amp running through that 68k resistor giving it an instantaneous power of ~0.5^2 x 68000 = ~17kW. The 3W resistor survives because 17kW for one us is 17mJ. The resistor can handle a 17mJ impulse. In doing so, it slows the ramp-rate of the coupling capacitor's connection to the inverter, thus protects it from receiving a sudden voltage sufficient to cause the inverter internal damage.

[ab0032]

You can buy the inverters cheaper at a place like mouser.com

Here is just one example of the price for 40 TDK brand inverters that output 2x 6mA at 31.16 Euros, they have a huge selection of inverters, dim-able and whatever. Most likely you will have to add something for shipping. Maybe they only sell to businesses, I dont know. Pick the country closest to you from mouser...

http://de.mouser.com/ProductDetail/TDK/ ... Ed3HXyQA6O

The dimmer function may be interesting for alternative circuit designs.

> I bought inverters from this site;
>
> http://shop.wiltec.info/product_info.ph ... -40mm.html
>
> Unfortunately, the price has gone up since I bought 100 at E1.25 each.

[Chris Bradley]

Hi Alex.

When I said 'the price has gone up', it hasn't gone up much! You can still buy 50 of the Wiltec items at 1.39 each (including taxes). So the whole purchase price of these is cheaper than a quarter of the additional VAT (alone) on the one you mentioned!

But you mention something I should have emphasised - most cheap inverters are intrinsically 'dimmable' because they simply pass the applied volts to the transformer within it, and so the output is 'analogue'. Whether the driver electronics work at variable volts may be a different question.

The ones I mentioned from Wiltec operate from just a volt or two up - I guess this is just the VBE of the internal bipolars that drive the simple oscillator circuit within. So you adjust the input volts to get control of the output power.

The only figures of merit worth worrying about are therefore the VBE and VCEO of the internal transistors that determine the lower and upper supply voltage limits. I'd hazard a guess that the ones in the Wiltec units are around 1.5V and 30V so volts inputs are OK but I think I have blown them up before in 'performance testing' due to excess current beyond their ICM.

It'd, obviously, be much less useful if you were stuck to one output power if your inverters could only operate at one given voltage, so you don't want ones like that if you are trying to build this circuit for a variable supply, but I don't think you'll find many like that in the really cheap varieties.

[Deiter Hanbicki]

Hi Chris,
The power inverters on Wiltech dont seem to exist anymore and I can't seem to find any that arent >$30, am I searching for the wrong thing or do cheap inverters not exist anymore?

[Chris Bradley]

Unfortunately, I can imagine a time when CCFL inverters will, indeed, no longer be available.

I guess wiltech have moved over to LEDs only.

For now you can still get CCFL inverters for 'old' computer monitors, but these are moving to LEDs too, so that supply will die out.
see... http://www.ebay.co.uk/itm/310952320305

At some stage you will have to go looking for ferrite transformers yourself and build your own oscillator. This is really not a very big challenge, but finding the transformers might be. (I've always been keen to get a hold of piezo-transformers, which would really suit this topology.)

Lots of alternative low-voltage technologies have been taking over high-voltage solutions since transistors took over from thermionic valves. That trend will continue I am sure, and loss of a cheap supply of CCFL inverters will eventually become loss of a supply of any CCFL inverters. Folks late to amateur plasma experiments will find, and are finding, that a supply of HV components, from resistors to capacitors and diodes, are simply drying up.

Things can only get more expensive, if available at all, as any real industrial demand drops away to zero. It seems these are twilight years for any sort of mass-produced high voltage parts.

[Richard Hull]

I will be very curious to see if any one using this topology ever enters the neutron club. This is a great idea and effort. All the best on this to any and all using it in a fusor. Slick thinking and design.

Richard Hull

[Benjamin Walsh]

Chris, I was wondering if you might be willing to part with a few of your inverters (for a price, of course)? It would be greatly appreciated.

Thanks for the great information and idea regardless of your decision.
-Ben


(Sorry to make a public post, but I lurk too much and don't post enough to have private messaging enabled. Oops.)

[Chris Bradley]

They are boxed up at the moment but in principle I could do that (if I can find them!!). What is the specification/project you are aiming for and how many do you want, where are you?

These days, it might be easier to find a smaller number of higher power lighting or OBIT transformers that are also useful in this arrangement, as they can make use of the 'GFI' type devices.

[Benjamin Walsh]

I was hoping to achieve 30kv at around 15 mA. I was looking at purchasing about 25-30 of your inverters. I live in Southern California (if you want to pm me your email, I can send you my address) about 30 minutes outside Los Angeles.
I hadn't considered an OBIT or similar transformer, and a cursory glance shows that it might end up being a tad more expensive than a CCFL based transformer, depending on how much you are willing to part with your inverters for. But a great option nonetheless!

Thanks,
Ben

[James Hammond]

Chris, how exactly did you go about powering the CCFL inverters you had?

[Richard Hull]

I build GM counters for sale at hamfests and use only CCFL inverters to obtain the HV. They typically take 5-12 volts DC, depending on the inverter and demand from 20ma to over 250ma of current at these voltages, again, depending on the inverter. Powering 1 to 50 inverters is a snap as construction of a low voltage, high current supply is child's play.

As my inverters need only 700 volts HV and only 100 microamps or so of current, the inverters I use run off 5 volts DC. They draw about 25ma from the 9 volt transistor battery I use in the GM counter. I use a LDO 5 volt regulator to power the inverter from the 9 volt battery. (Lets the battery voltage drop pretty low before it needs replacement.)

I also have some very small Toshiba inverters that can supply 1.3kv @ 3ma. These run off 12 volts and drag 370ma from the supply!

Richard Hull

[Benjamin Walsh]

I'm just going to leave this here.... You can get 5 (absolutely 100% positively free. With pretty fast shipping too!) samples for 8 different inverters.

http://www.coilcraft.com/ccfl.cfm

[Dave Xanatos]

I'm just going to leave this here.... You can get 5 (absolutely 100% positively free. With pretty fast shipping too!) samples for 8 different inverters.

http://www.coilcraft.com/ccfl.cfm

Cute little transformers, but how would you ever get enough of them to generate the CURRENT you need in a fusor? You'd need like a hundred of these

[Benjamin Walsh]

Seeing as this ( https://web.archive.org/web/20121018042 ... -40mm.html ) is what Chris used, I think that somewhat less than 100 might work.... 2 of the inverters (for 10 free samples between the two) are rated higher than Chris's, so hopefully they can be run at higher voltage/current like Chris did to his. He ended up using 6 inverters to make a 30kv 60 watt transformer. And even if these end up not living up to quite what his were, there are still extras that can be thrown into the circuit to boost the current. And it'll cost you pretty much nothing. For freebies I'm not going to complain too much.

[Richard Hull]

The hottest transformer I saw was 2kv @11ma that means 15 would make 30 kv in series with zero current reserve. (The foregoing would demand a primary drive of about 400 watts of high frequency energy assuming about 80% conversion efficiency!) That 80 watts is long gone and exceeded.

I would wonder how well the secondaries might connect in parallel? This would be demanded to give more useful current.
Isolation on the primary driver systems might become an issue especially if driven in parallel. Lots to work out here, but maybe worth it. Good luck to all who venture out here in an attempt to get real, viable fusion voltage and current in fusion load and operational conditions.

Richard Hull